Discussion
Empirical models
The evaluation of swell behavior of a soil using undisturbed samples and specialized swell
tests is a difficult and expensive process for practicing engineers and small builders.
Therefore, there is a need for simple routine tests that can be performed on disturbed
engineered samples to achieve the same purposes. The empirical models appearing in the
literature are primarily related to prediction of swelling and swelling pressure from index
properties of soils. Sometimes, the empirical models proposed cannot be applied appropriately
to all soils due to different soil conditions and testing procedures.
The data used, here, was compiled from different determinations of swelling and index
properties. It was hoped that the models would be acceptable and generalized. Figure 3.6
shows the correlation between predicted-swelling and actual (measured) swelling. The figure
plotted all the data used (i.e. 115 data samples were used in learning and validating). The
dashed line shows the correlation between measured and predicted swell. It was expected
that correlation should have lain on the 45 line (1 : 1 line), which refers to the colinearity
of model. The figure illustrates that proposed empirical models, given in Table 3.3, give
good correlations. The dash lines in Figures 3.6a, 3.6b, and 3.6c were laid down in the colinearity
range of 0.5–0.8. The empirical model, proposed by Seed et al. (1962), as shown in
Figure 3.6d, showed a very weak correlation in which the correlation was below line 0.5.
It indicated that the proposed equation by Seed et al. (1962) is only appropriate for a
measured-swelling of less than 30%. In our study, the multiple linear regression method
(Equation 3.4, in Table 3.4) indicated a best-fit correlation. In general, the model can be
used for all soil conditions.
In the current study, multiple regression analysis was considered to derive an equation
that can be used to predict swelling potential from several index and physical properties. The
Discussion
Empirical models
The evaluation of swell behavior of a soil using undisturbed samples and specialized swell
tests is a difficult and expensive process for practicing engineers and small builders.
Therefore, there is a need for simple routine tests that can be performed on disturbed
engineered samples to achieve the same purposes. The empirical models appearing in the
literature are primarily related to prediction of swelling and swelling pressure from index
properties of soils. Sometimes, the empirical models proposed cannot be applied appropriately
to all soils due to different soil conditions and testing procedures.
The data used, here, was compiled from different determinations of swelling and index
properties. It was hoped that the models would be acceptable and generalized. Figure 3.6
shows the correlation between predicted-swelling and actual (measured) swelling. The figure
plotted all the data used (i.e. 115 data samples were used in learning and validating). The
dashed line shows the correlation between measured and predicted swell. It was expected
that correlation should have lain on the 45 line (1 : 1 line), which refers to the colinearity
of model. The figure illustrates that proposed empirical models, given in Table 3.3, give
good correlations. The dash lines in Figures 3.6a, 3.6b, and 3.6c were laid down in the colinearity
range of 0.5–0.8. The empirical model, proposed by Seed et al. (1962), as shown in
Figure 3.6d, showed a very weak correlation in which the correlation was below line 0.5.
It indicated that the proposed equation by Seed et al. (1962) is only appropriate for a
measured-swelling of less than 30%. In our study, the multiple linear regression method
(Equation 3.4, in Table 3.4) indicated a best-fit correlation. In general, the model can be
used for all soil conditions.
In the current study, multiple regression analysis was considered to derive an equation
that can be used to predict swelling potential from several index and physical properties. The
การแปล กรุณารอสักครู่..
![](//thimg.ilovetranslation.com/pic/loading_3.gif?v=b9814dd30c1d7c59_8619)